In 5,209 subjects studied for 24 years in Framingham, Mass, 691 cases of cancer were documented, with histological confirmation for 94%. Predetermined personal characteristics were tested for associations with subsequent occurrence of cancer at specific sites using multiple logistic regression. Significant associations of various cancer sites with cigarette smoking, alcohol use, education, height, weight, and parity agreed with other studies. Serum cholesterol level was inversely associated with incidence of colon cancer and with other sites only in men; these inverse associations were statistically significant after adjustment for age, alcohol consumption, cigarette smoking, education, systolic blood pressure, and relative weight. Associations may reflect effects of competing lethal diseases, underlying pathophysiological mechanisms that promote or inhibit development of cancer in men, biologic or social response to early and undiagnosed states of cancer.